Record:   Prev Next
Author Holder, Eugene Michel, II
Title Quadrilateral mesh smoothing using a genetic algorithm
book jacket
Descript 210 p
Note Source: Dissertation Abstracts International, Volume: 63-01, Section: B, page: 0402
Chair: Charles L. Karr
Thesis (Ph.D.)--The University of Alabama, 2001
This dissertation presents the results of an investigation in which genetic algorithms were used for general 2-D quadrilateral mesh smoothing, using one or multiple nodes, in finite element analysis (FEA). Specifically, a useful engineering tool was developed for achieving 2-D quadrilateral mesh smoothing. The tool is called the genetic algorithm smoother (GAS). To improve the operation of GAS, a technique termed feasible circles was developed to reduce the size of the requisite search areas, thereby dramatically improving genetic algorithm performance. GAS can be used to achieve either pseudo-meshing (predefined connectivity) or mesh untangling of up to 25 elements (16 nodes). In addition, GAS can simultaneously manipulate up to 64 nodes, with increasingly long computer run times as the problem size increases
Five variations of GAS (GAS-1, GAS-4, GAS-9, GAS-N, and GAS-M) are presented with the associated results of using a variety of FEA models. The results of these experiments are compared to results achieved using the Laplace smoothing technique. These results indicate that GAS is a reasonably effective tool for solving the problem of 2-D mesh smoothing
GAS requires that a measure of solution quality be defined by the user. In this effort, a distortion metric was used to quantify the "goodness" of individual quadrilateral elements. The distortion metric was combined with weighted measurements of the interior angles and the aspect ratio of elements to form an objective function for GAS. Other implementation details, such as the convergence criteria, population size, crossover probability, and mutation rate, are presented in the dissertation
Parametric studies designed to test the effectiveness of GAS were performed, and the results are presented in both tabular and graphical form. Optimal genetic algorithm parameters are listed for the mesh-smoothing process using a steady-state genetic algorithm. Planned future work is outlined and possible genetic algorithm smoothing applications are discussed
School code: 0004
Host Item Dissertation Abstracts International 63-01B
Subject Engineering, Civil
Engineering, Mechanical
Engineering, Aerospace
Artificial Intelligence
Alt Author The University of Alabama
Record:   Prev Next